In-app assistant
Basemind includes an AI assistant (Kuma) in the app itself. It uses the same tools as external MCP agents, so it can explain your data, build tables, write formulas, create applications and dashboards, and work with project memory.
Enabling the assistant
The assistant needs an AI provider:
- As instance administrator open Admin → AI providers and add a provider (OpenAI or an OpenAI-compatible endpoint, Anthropic, AWS Bedrock, Groq, Google Gemini, Ollama…) with its credentials and models.
- On each model, choose what it may be used for: Kuma (the assistant), AI fields and/or AI agent actions. Click Test model to check it.
- In the AI features section select the model Kuma uses.
- Each workspace can inherit that choice, pick another model or disable the assistant in its own AI provider settings.
Without database-configured providers the server falls back to the
BASEROW_ENTERPRISE_ASSISTANT_LLM_MODEL environment variable (set it on both
the backend and the web-frontend). See Self-hosting.
Using it
- Open the assistant panel from the sidebar and type your request.
- On a project page the assistant knows which project you are on. The project’s Overview tab also has an Ask the assistant box.
- The assistant shows which tools it calls and links to what it created.
Example prompts:
- “Explain the formula in the Margin field.”
- “Add a Priority single select with Low, Medium and High to Tasks.”
- “Summarise the decisions we made in this project last month.”
- “Create a form page where clients can submit support requests.”
Modes and safety
- In explain mode the assistant only reads.
- In database mode it may also write; inside a project its writes follow the project’s autonomy level and appear in the agent activity log, exactly like MCP calls.
- The in-app assistant never has Full access.
AI fields and formulas
Independently of the assistant, the configured AI provider also powers the AI field (a field that generates text per row from a prompt) and the generate formula helper in the formula editor.